Re: Re: Replacing $mon with Jan, Feb, Mar by regex or other means

I know you probably weren't serious.

But it turns out that your approach isn't as slow as I thought it would be. It's still slower than using an array, but not terribly.

Benchmark: running inlineArrayLookup, preBuiltArrayLookup, strftimeLoo
+kup, substrLookup, each for at least 2 CPU seconds...
inlineArrayLookup:  2 wallclock secs ( 2.12 usr + -0.01 sys =  2.11 CP
+U) @ 543540.28/s (n=1146870)
preBuiltArrayLookup:  1 wallclock secs ( 2.04 usr + -0.01 sys =  2.03 
+CPU) @ 588310.34/s (n=1194270)
strftimeLookup:  3 wallclock secs ( 1.76 usr +  0.37 sys =  2.13 CPU) 
+@ 216666.20/s (n=461499)
substrLookup:  2 wallclock secs ( 2.09 usr +  0.00 sys =  2.09 CPU) @ 
+465374.16/s (n=972632)
                        Rate strftimeLookup substrLookup inlineArrayLo
+okup preBuiltArrayLookup
strftimeLookup      216666/s             --         -53%              
+-60%                -63%
substrLookup        465374/s           115%           --              
+-14%                -21%
inlineArrayLookup   543540/s           151%          17%              
+  --                 -8%
preBuiltArrayLookup 588310/s           172%          26%              
+  8%                  --
[download]

Comparing the strftime version isn't totally fair, since it's not a simple "month->string" converter, but requires a localtime list. But as others have pointed out, it will do locales correctly.

Here's the benchmark code:

#!/usr/bin/perl -w

use strict;
use Benchmark qw(cmpthese);
use POSIX 'strftime';

my @months=qw(Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec);

sub getMonthNumber {
    return int(rand(12));
}

sub preBuiltArrayLookup {
    return $months[getMonthNumber()];
}

sub inlineArrayLookup {
    return qw(Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec)[getMont
+hNumber()];
}

sub substrLookup {
    return substr('JanFebMarAprMayJunJulAugSepOctNovDec',getMonthNumbe
+r()*3,3);
}

# so as to not penalize strftime with localtime call
my @fixedLocaltime=localtime;
sub strftimeLookup {
    return strftime "%m",@fixedLocaltime;
}

cmpthese(-2,
        {
             'preBuiltArrayLookup' => \&preBuiltArrayLookup,
             'inlineArrayLookup' => \&inlineArrayLookup,
             'substrLookup'   => \&substrLookup,
             'strftimeLookup'   => \&strftimeLookup,
         }
     );
[download]

--
Mike

Edit: Added strftime variant which I had overlooked.

Edit by tye, change PRE to CODE around long lines

Comment on Re: Re: Replacing $mon with Jan, Feb, Mar by regex or other means Select or Download Code

Replies are listed 'Best First'.
Re: Replacing $mon with Jan, Feb, Mar by regex or other means by Abigail-II (Bishop) on Apr 25, 2003 at 12:06 UTC
To make a better comparison, you should `strftimeLookup` do a `getMonthNumber` as well. Now it's the only case that isn't penalized by an extra subroutine call. Or you could eliminate the call from all cases, the lookup time won't be different depending on the number of the month. Just use a fixed month number (but don't hardcode it as a literal, or else the compiler might pre-calculate some calculations). Abigail	[reply] [d/l] [select]
Re: Re: Replacing $mon with Jan, Feb, Mar by regex or other means by RMGir (Prior) on Apr 25, 2003 at 12:14 UTC
I did consider that, but since the strftime variant was already losing badly, it didn't seem critical. You're right, though, wrapping the fetch of @fixedLocaltime in a sub (returning a ref to the array) call kills the performance of the strftime variant a bit more. $ perl testMonth.pl Benchmark: running inlineArrayLookup, preBuiltArrayLookup, strftimeLoo +kup, strftimeSubLookup, substrLookup, each for at least 2 CPU seconds +... inlineArrayLookup: 2 wallclock secs ( 2.12 usr + 0.00 sys = 2.12 CP +U) @ 513847.17/s (n=1089356) preBuiltArrayLookup: 2 wallclock secs ( 2.09 usr + 0.01 sys = 2.10 +CPU) @ 568700.00/s (n=1194270) strftimeLookup: 1 wallclock secs ( 1.81 usr + 0.35 sys = 2.16 CPU) +@ 223003.24/s (n=481687) strftimeSubLookup: 2 wallclock secs ( 1.74 usr + 0.26 sys = 2.00 CP +U) @ 182455.50/s (n=364911) substrLookup: 2 wallclock secs ( 2.19 usr + 0.01 sys = 2.20 CPU) @ +446835.91/s (n=983039) Rate strftimeSubLookup strftimeLookup substrLo +okup inlineArrayLookup preBuiltArrayLookup strftimeSubLookup 182455/s -- -18% +-59% -64% -68% strftimeLookup 223003/s 22% -- +-50% -57% -61% substrLookup 446836/s 145% 100% + -- -13% -21% inlineArrayLookup 513847/s 182% 130% + 15% -- -10% preBuiltArrayLookup 568700/s 212% 155% + 27% 11% -- [download] -- Mike Edit by tye, change PRE to CODE around long lines	[reply] [d/l]
Re: Re: Re: Replacing $mon with Jan, Feb, Mar by regex or other means by BrowserUk (Patriarch) on Apr 25, 2003 at 12:49 UTC
I do use this, though I usually put the constant information into a constant. `use constant MONTHS=> 'JanFebMarAprMayJunJulAugSepOctNovDec';` [download] You'll find that this has little if any effect on the performance. Examine what is said, not who speaks. 1) When a distinguished but elderly scientist states that something is possible, he is almost certainly right. When he states that something is impossible, he is very probably wrong. 2) The only way of discovering the limits of the possible is to venture a little way past them into the impossible 3) Any sufficiently advanced technology is indistinguishable from magic. Arthur C. Clarke.	[reply] [d/l]